Abstract
The usual theory of prediction with expert advice does not differentiate between good and bad “experts”: its typical results only assert that it is possible to efficiently merge not too extensive pools of experts, no matter how good or how bad they are. On the other hand, it is natural to expect that good experts' predictions will in some way agree with the actual outcomes (e.g., they will be accurate on the average). In this paper we show that, in the case of the Brier prediction game (also known as the quadratic-loss game), the predictions of a good (in some weak and natural sense) expert must satisfy the law of large numbers (both strong and weak) and the law of the iterated logarithm; we also show that two good experts' predictions must be in asymptotic agreement. Finally, we briefly discuss possible extensions of our results to more general games; the limit theorems for sequences of events in conventional probability theory correspond to the log-loss game.
This is a preview of subscription content, log in via an institution.
Preview
Unable to display preview. Download preview PDF.
References
P. Billingsley, Probability and measure, New York: Wiley, 1979.
N. Cesar-Bianchi, P. Long, and M. K. Warmuth, Worst-case quadratic loss bounds for on-line prediction of linear functions by gradient descent, IEEE Trans. Neural Networks, 7 (1996), 604–619.
T. Cover and E. Ordentlich, Universal portfolios with side information, IEEE Trans. Inform. Theory, 42 (1996), No. 2.
A. P. Dawid, Calibration-based empirical probability (with discussion), Ann. Statist., 13 (1985), 1251–1285.
A. P. Dawid and V. G. Vovk, Prequential probability: principles and properties, submitted to Bernoulli.
A. DeSantis, G. Markowsky, and M. N. Wegman, Learning probabilistic prediction functions, in “Proceedings 29th Annual IEEE Symposium on Foundations of Computer Science”, pp. 110–119, Los Alamitos, CA: IEEE Comput. Soc., 1988.
D. Haussler, J. Kivinen, and M. K. Warmuth, Tight worst-case loss bounds for predicting with expert advice, Technical Report UCSC-CRL-94-36, revised December 1994. Short version in “Computational Learning Theory” (P. Vitányi, Ed.), Lecture Notes in Computer Science, Vol. 904, pp. 69–83, Berlin: Springer, 1995.
M. Herbster and M. K. Warmuth, Tracking the best expert, in “Proceedings, 12th International Conference on Machine Learning”, 1995.
J. Kivinen and M. K. Warmuth, Additive versus exponentiated gradient updates for linear prediction, Inform. Computation, 132 (1997), 1–64.
N. Littlestone, P. M. Long, and M. K. Warmuth, On-line learning of linear functions, J. Computational Complexity, 5 (1995), 1–23.
N. Littlestone and M. K. Warmuth, The Weighted Majority Algorithm, Inform. Computation, 108 (1994), 212–261.
G. Shafer, The art of causal conjecture, Cambridge, MA: MIT Press, 1996.
R. J. Solomonoff, Complexity-based induction systems: comparison and convergence theorems, IEEE Trans. Inform. Theory, IT-24 (1978), 422–432.
V. Vovk, On a randomness criterion. Soviet Math. Doklady 35 (1987), 656–660.
V. Vovk, Aggregating strategies, in “Proceedings 3rd Annual Workshop on Computational Learning Theory” (M. Fulk and J. Case, Eds.), pp. 371–383, San Mateo, CA: Morgan Kaufmann, 1990.
V. Vovk, A logic of probability, with application to the foundations of statistics (with discussion), J. R. Statist. Soc. B, 55 (1993), 317–351.
V. Vovk, A game of prediction with expert advice, to appear in J. Comput. Inform. Syst. Short version in “Proceedings, 8th Annual ACM Conference on Computer tional Learning Theory”, pp. 51–60, New York: Assoc. Comput. Mach., 1995.
V. Vovk, Derandomizing stochastic prediction strategies, in “Proceedings, 10th Annual ACM Conference on Computational Learning Theory”, to appear.
V. Vovk, On-line competitive linear regression, Technical Report CSD-TR-97-13, Department of Computer Science, Royal Holloway, University of London, 1997.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1997 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Vovk, V. (1997). Probability theory for the Brier game. In: Li, M., Maruoka, A. (eds) Algorithmic Learning Theory. ALT 1997. Lecture Notes in Computer Science, vol 1316. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-63577-7_52
Download citation
DOI: https://doi.org/10.1007/3-540-63577-7_52
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-63577-2
Online ISBN: 978-3-540-69602-5
eBook Packages: Springer Book Archive